Skip to content
Try Free →

Why we built AskVault

Last updated: · 5 min read

The three broken patterns

In B2B SaaS, customer support has three patterns that don't work:

Pattern 1: customer can't find the answer. Your docs exist. Customers don't find them. They open a ticket for something that was already documented.

Pattern 2: agents waste time on repetitive questions. The same order-status question, the same refund-policy question, again and again. Skilled agents do entry-level work.

Pattern 3: existing bots hallucinate. Flow-builder bots can't handle unscripted questions. RAG bots without grounding make up facts. Customers stop trusting bots after one bad answer.

What we set out to build

The opposite of those patterns:

  • Customers find answers fast. Bot answers straight from your docs.
  • Agents focus on hard cases. Bot handles repetitive; humans handle real escalations.
  • Bot answers are source-cited. Each answer shows the sources it was drawn from (on by default).
  • Grounded answers. The model is told to answer only from retrieved content. That reduces hallucination; it does not eliminate it.
  • Self-serve setup. From your website URL to an embeddable widget, no sales call.
  • Native multi-channel. One agent answers on your website widget, WhatsApp, Slack, Discord, Telegram, email and more.

Why source citations matter

A bot that says "Our refund policy is 14 days" without citation is unverifiable. Customers don't know if it's true. Agents have to double-check. Trust erodes.

A bot that says "Per the Refund Policy, refunds are 14 days. Last updated May 1." is verifiable. Customers can check it, and so can your agents.

That is why the source list is on by default. Each workspace can turn it off on its Playground page.

How pricing works

Each plan is metered on two things, and both are listed up front:

  • Monthly message credits. Each answer uses credits: base models use 1, premium models use more. Credits are shared across all your channels.
  • Indexed content (MB). A cap on how much content your agents can be trained on.

See pricing breakdown and why MB-cap.

What we're not

Three honest exclusions:

  • Not a general-purpose chatbot. We're optimized for customer support and helpdesk.
  • Not a flow-builder. We're RAG-first.
  • Not the cheapest. Cheaper alternatives exist.

What we focus on

  • Native multi-channel. Train once, then deploy the same agent on your web widget, WhatsApp, Slack, Discord, Telegram, email and more.
  • Transparent pricing. Each plan lists its monthly message credits and indexed-content (MB) cap up front.
  • Answer quality you can trust. Hybrid retrieval with reranking and source-cited answers, not free-form generation.
  • Agentic skills. The bot can take actions, not just retrieve text.
  • Self-serve setup. From your website URL to a live widget without a sales call.
  • A database provider with a SOC 2 Type II report. AskVault has no SOC 2 report of its own, and there is no BAA (HIPAA).

Plans are priced in USD.

Where we started

When we started, Chatbase was the obvious comparison. We respected their product and set out to build our own take: RAG-first, sources shown with answers, and one agent across many channels. For a current side-by-side, check each vendor's own pricing page; plans and prices change.

The roadmap mindset

We treat documentation as commitments:

  • Features we have are documented in detail.
  • Features we plan are marked as roadmap.
  • Features we don't have are explicit.

If you find a page that describes something the product does not do, tell us and we will fix the page.

Values

Three core values:

  1. Trust through transparency. Source citations and clear pricing.
  2. Up-front billing. Prices, message credits and content caps are listed before you pay.
  3. Build for the long term. Not a quick exit.

How to reach us

Limits

  • What we promise. Source-cited answers and transparent pricing.
  • What we don't promise. Perfect bot, zero hallucination, all features today.

FAQ

Will you be acquired?

Long-term play; no acquisition plans.

Was this page helpful?